OpenClaw cut CI runner-hours by 86.5% and compute value by 95.5%: here's how
vincent_koc · x · 2026-11-17
Vincent Koc's writeup on how OpenClaw hit a CI bottleneck from coding agents producing more tests, jobs and retries. Weekly Blacksmith runner usage fell from 158,546 hours (September peak) to 21,395 hours — an 86.5% drop — with estimated compute value down 95.5% (capacity was sponsored; workload also changed).
Key moves:
- Remove pointless checks and stop repeating expensive setup
- Make better use of each machine
- Adaptive test selection: fall back to full suite when affected-test selection is uncertain; keep required checks
- 89% of the codebase is TypeScript, which is where faster tools landed first
The post ships a copy-paste prompt to hand your agent: audit your repo's CI, measure a baseline, propose and implement high-value changes in reviewable steps, then verify correctness, feedback time and compute cost.
More from coding & agent
- Prompt of the day: have your agent study all skills and build a skill orchestrator — yungcontent · 2026-11-17
- Databricks benchmark: Genie passes 91% of enterprise data tasks vs 45–58% for frontier agents — matei_zaharia · 2026-11-17
- Financial Datasets launches Stock Screener API: 17,000 stocks, from $0.02 via x402 — virattt · 2026-11-17
- Stock Screener API launches on x402: agents screen 17,000 stocks for $0.02 — virattt · 2026-11-17
- Dev builds an "omni-feed" event bus to monitor all agent actions in one pane — DavidWells · 2026-11-17
- Dev calls for an 'omni-feed' product: one filterable pane for all agent actions — DavidWells · 2026-11-17